Looking for 5 AI-agent teams that need durable cross-session memory

[decebal]

I’m opening five design-partner spots for builders working on real AI-agent systems.

The problem I want to work on together: agent memory that survives restarts, preserves where a recalled fact came from, and can reconstruct what the agent knew before a later correction.

Each selected team gets:

• 60 days of hosted AllSource Scale access
• a founder-led integration session
• direct engineering help through one working memory flow
• two focused feedback calls

Good fit: production or serious pre-production agents where stale summaries, missing provenance, or lost cross-session state already create engineering cost. You should be able to integrate within 60 days.

This is product research, not a review program. No endorsement, testimonial, star, or public post is required.

Apply in about three minutes:
https://www.all-source.xyz/design-partners?utm_source=github&utm_medium=community&utm_campaign=design_partners_2026&utm_content=discussion

I’m also glad to answer technical questions here about event-sourced agent memory, Core, or Prime before anyone applies.